A Study of Pattern-Based Suptopic Discovery and Integration in the Web Track.
Wei Zheng, Hui Min Fang · 2011
We report our systems and experiments in the diversity task of TREC 2011 Web track. Our goal is to evaluate the effectiveness of the proposed methods for subtopic extraction and diversification steps on the large data collection. In the subtopic extraction step, we extract subtopics using both structured data, i.e., ODP, which provides high quality information and unstructured data, i.e., original retrieved documents, which contains terms effective in diversifying documents. In the diversification step, we use a coverage-based method to diversify documents based on the extracted subtopics. It has the desired properties of diversification which favors documents covering more subtopics and documents covering novel subtopics that have not been well covered by previously selected documents.